IFA/MMS GA shift optimization refactor

[tylerdy]

IFA/MMS GA shift optimization refactor

Description

• Category: artifact
• JIRA issue: https://jira.ihme.washington.edu/browse/SSCI-2509
• Research reference: https://ihme.slack.com/archives/C02944X688L/p1765571463092489

Changes and notes

Stratify IFA GA shift optimization by ANC attendance for IFA calibration refactor, and update MMS to use ANC exposure data.

Update artifact to use new optimization outputs. See comment below for the changes to artifact key values.

New artifacts at /mnt/team/simulation_science/pub/models/vivarium_gates_mncnh/artifacts/ifa_mms_refactor/

Verification and Testing

See IFA and MMS validations after optimization cells, and the comment linked above to see what values changed in the artifact.

*** REMINDER ***
CI WILL NOT RUN ANY TESTS.
MANUALLY RUN TESTS WITH EACH PR.
-->

• all tests pass (pytest --runslow)

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[zmbc]

Where did this file come from?

[alibow]

Yay!! Looks great! I say it's ready to merge :) thanks Tyler!

[tylerdy]

@hussain-jafari I've updated the artifact based on the changes to the GA shift calibration. See the changes to artifact key values in this notebook with the below parameters. @alibow your review of the key changes may be helpful as well.

old_art_dir = "/mnt/team/simulation_science/pub/models/vivarium_gates_mncnh/artifacts/model25.0/"
new_art_dir = "/mnt/team/simulation_science/pub/models/vivarium_gates_mncnh/artifacts/ifa_mms_refactor/"
loc = "ethiopia"

One question (probably for @hussain-jafari) I have about the key changes is whether its important for 'risk_factor.iron_folic_acid_supplementation.risk_specific_shift' to contain both GA and BW shifts. Right now it only contains GA, where as 'risk_factor.multiple_micronutrient_supplementation.risk_specific_shift' only contains BW. The IFA risk specific shift used to be calculated from the excess shift key which contained both GA and BW shifts, but now is calculated from the weighted average GA shift. If its important to add back the BW shift I can do that, but I wasn't sure what that would accomplish.

[tylerdy]

I updated the artifact so that the keys can be used with the existing model code. See the updated diff for the artifact keys here

[tylerdy]

Tests are passing with the data updates. Renamed artifacts from

/mnt/team/simulation_science/pub/models/vivarium_gates_mncnh/artifacts/ifa_mms_refactor/
to
/mnt/team/simulation_science/pub/models/vivarium_gates_mncnh/artifacts/model30.0

[zmbc]

Looks like we're continuing to exaggerate this effect a bit, but it's not worse than in the previous run.

Reply via ReviewNB